feat(api): add detailer postprocess script and /sdapi/v1/detail endpoint

Surface YoloRestorer.restore() as a standalone operation: a Detailer
postprocessing script in the Process tab and a thin /sdapi/v1/detail
endpoint, neither requiring a base generation pass.

- modules/postprocess/yolo.py: YoloRestorer.make_processing() builds the
  synthetic Img2Img processing object both entry points feed to restore(),
  resolving the seed so the inpaint passes are reproducible
- modules/api/process.py: post_detail handler exposes the full detailer
  parameter set and returns the detailed image plus optional annotations
  as base64
- scripts/postprocessing_detailer.py: reuses shared.yolo.ui('extras') and
  runs through make_processing()
- modules/postprocessing.py: run_extras takes a per-script script_args
  dict, also letting the extras API drive other scripts such as Remove
  background; omitting it leaves existing callers unchanged
- modules/api/models.py: ReqDetail / ResDetail
- modules/processing_info.py: guard create_infotext's Image/Hires CFG
  reporting against an unset (None) cfg_image, matching the is-not-None
  checks the other cfg_image readers use; the detailer inpaint pass runs
  with it unset
- test/test-detailer-api.py: covers both paths; effect tests measure the
  diff inside the detected region with extreme isolated parameter values,
  and the suite disables model quantization for the run and restores the
  original settings afterward
This commit is contained in:
CalamitousFelicitousness
2026-05-07 22:47:43 +01:00
parent c3eec8e60b
commit f9ab0bf04d
9 changed files with 715 additions and 56 deletions
+76
View File
@@ -0,0 +1,76 @@
import numpy as np
from PIL import Image
from modules import scripts_postprocessing, shared
from modules.logger import log
class ScriptPostprocessingDetailer(scripts_postprocessing.ScriptPostprocessing):
name = "Detailer"
order = 15000
def ui(self):
# The detailer accordion (built by yolo.ui) now contains the Sampler sub-accordion too, so for 'extras'
# it returns a 7th element: a dict of the sampler-block controls. Spread it into the control map; their
# values are stamped onto the synthetic p in process()/make_processing(), applying to this pass only.
enabled, prompt, negative, steps, strength, resolution, sampler_block = shared.yolo.ui('extras')
return {
"enabled": enabled,
"prompt": prompt,
"negative": negative,
"steps": steps,
"strength": strength,
"resolution": resolution,
**sampler_block,
}
def process(self, pp: scripts_postprocessing.PostprocessedImage, # pylint: disable=arguments-differ
enabled=False, prompt='', negative='', steps=10, strength=0.3, resolution=1024,
sampler='Default', prediction='default', shift=3.0, cfg_scale=6.0, options=None, seed=-1):
if not enabled:
return pp
if shared.sd_model is None or not hasattr(shared.sd_model, 'sd_checkpoint_info'):
log.warning('Detailer postprocess: no base model selected')
pp.info["Detailer"] = "skipped (no base model selected)"
return pp
# The sampler block is stamped onto the synthetic p. The schedulers_* values become per-job overrides in
# processing_helpers (they beat the global opts for this pass only); a named sampler is required for them
# to take effect, 'Default' keeps the model scheduler. cfg_scale and hr_sampler_name apply directly.
options = options or []
overrides = {
'hr_sampler_name': sampler,
'schedulers_prediction_type': prediction,
'schedulers_shift': shift,
'cfg_scale': cfg_scale,
'schedulers_use_loworder': 'low order' in options,
'schedulers_use_thresholding': 'thresholding' in options,
'schedulers_dynamic_shift': 'dynamic' in options,
'schedulers_rescale_betas': 'rescale' in options,
}
log.info(f'Detailer postprocess: strength={strength} steps={steps} resolution={resolution} sampler={sampler} cfg={cfg_scale}')
p = shared.yolo.make_processing(pp.image, prompt=prompt, negative=negative, steps=steps, strength=strength, resolution=resolution, seed=int(seed) if seed is not None else -1, overrides=overrides)
try:
result = shared.yolo.restore(np.array(pp.image), p)
except Exception as e:
log.error(f'Detailer postprocess: {e}')
return pp
# restore() returns list[ndarray] (detailed image at [0], annotated debug at [1] when enabled)
# on success, or a single ndarray on early-return paths. The postprocessing pipeline is one
# image per input, so the annotated debug image is dropped here; use /sdapi/v1/detail for it.
if isinstance(result, list) and len(result) > 0:
pp.image = Image.fromarray(result[0])
elif isinstance(result, np.ndarray):
pp.image = Image.fromarray(result)
pp.info["Detailer"] = "Enabled"
pp.info["Detailer strength"] = strength
pp.info["Detailer steps"] = steps
pp.info["Detailer resolution"] = resolution
pp.info["Detailer sampler"] = sampler
if prompt:
pp.info["Detailer prompt"] = prompt
if negative:
pp.info["Detailer negative"] = negative
return pp